While being the "manufacturer's own" compiler does have its marginal advantages, for Atmel Studio in this instance, the corresponding disadvantages are many:
- Less people will see and report bugs, or suggest improvements, than for a product with greater reach, such as IAR - Not just Atmel users buy IAR. So less fixes or enhancements will get released. Here the reference is to Atmel-specific bugs and features, not generic development environment ones.
- Development investment is more difficult to justify for a business (i.e. Atmel) whose primary business product is not the compiler. Hence, less features released per unit time.
- For a lot of development houses, the time spent in learning one more new interface is an unjustifiable expense, if an already familiar development platform can be used instead. This reduces the re-training cost to the portions specific to the target hardware architecture alone. IAR is multiplatform, hence a better investment in this sense.
- IAR specifically has useful features in the team / distributed development space, Microsoft's Visual Studio reserves its own such collaboration features for its enterprise versions, not the standalone version underlying Atmel Studio
Besides this, the question generically applies to all development platforms - Why would two or three or a dozen products commercially survive if there is one already existent that does the job?
For instance, I use Code Composer Studio whenever I have the option, even if I have access to IAR or Atmel Studio - not because CCS is definitively better in some specific way, but because it is what I am most comfortable with. Others similarly swear by their preferred flavor, just as with beverage preferences: "that subtle aroma of peat and that full-bodied richness from the seasoned casks".